... Resin Transfer Molding Process. On-Line Characterization of Bulk Permeability and Race-Tracking During the Filling ... Email Alerts: .... requires one to create a database using mold filling simulations that will contain the resin arrival times at ...
Journal of Composite Materials http://jcm.sagepub.com
On-Line Characterization of Bulk Permeability and Race-Tracking During the Filling Stage in Resin Transfer Molding Process Mathieu Devillard, Kuang-Ting Hsiao, Ali Gokce and Suresh G. Advani Journal of Composite Materials 2003; 37; 1525 DOI: 10.1177/0021998303034459 The online version of this article can be found at: http://jcm.sagepub.com/cgi/content/abstract/37/17/1525
Published by: http://www.sagepublications.com
On behalf of: American Society for Composites
Additional services and information for Journal of Composite Materials can be found at: Email Alerts: http://jcm.sagepub.com/cgi/alerts Subscriptions: http://jcm.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations (this article cites 19 articles hosted on the SAGE Journals Online and HighWire Press platforms): http://jcm.sagepub.com/cgi/content/refs/37/17/1525
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
On-line Characterization of Bulk Permeability and Race-tracking During the Filling Stage in Resin Transfer Molding Process MATHIEU DEVILLARD,* KUANG-TING HSIAO, ALI GOKCE AND SURESH G. ADVANI Center for Composite Materials and Department of Mechanical Engineering, University of Delaware, Newark, DE 19716, USA (Received August 21, 2002) (Revised January 13, 2003)
ABSTRACT: Resin Transfer Molding (RTM) is widely used to manufacture polymer composite materials. In this process, the fiber preform is placed in a closed mold and thermoset resin is injected to saturate the preform. After the resin cures the mold is opened and the net shape composite part is obtained. With RTM, one is capable of making complex and high quality composite parts with short cycle times. However, by introducing more complexity into the part, one also introduces higher probability of disturbances, such as race-tracking of resin during impregnation along preform edges. This can lead to incomplete saturation of fiber preform forming flaws such as dry spots in the composite part. The strength and existence of race-tracking is a function of the fabric type, perform manufacturing method, and their placement in the mold. It can vary from one part to the next in the same production run and usually it is not repeatable. The characterization of race-tracking and preform permeability is a key input for simulation tools that optimize the gates and vents locations and for designing advanced sensing and flow control strategies. This paper describes a method to estimate the strength and location of race-tracking and characterize the preform bulk permeability during the resin impregnation stage of RTM. The approach can be described in two steps. The first step generates simulations off line that map the overall range of potential disturbances likely to occur during the injection. The second step uses sensors placed in the mold, which allow one to identify the simulation within the database that resembles the experiment in terms of the filling behavior during the filling stage on the manufacturing platform. In the final step, this information is used to estimate race-tracking strength and location as well as the bulk permeability. Both experimental and numerical case studies show the reliability and accuracy of this method. It could be utilized for a broad range of applications such as active flow control, vent location optimization, and process monitoring of mold filling processes.
*Author to whom correspondence should be addressed.
Journal of COMPOSITE MATERIALS, Vol. 37, No. 17/2003 0021-9983/03/17 1525–17 $10.00/0 DOI: 10.1177/002199803034459 ß 2003 Sage Publications Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1525
1526
M. DEVILLARD
ET AL.
KEY WORDS: advanced composites, automation, process control, process simulation, resin transfer molding (RTM), race-tracking characterization.
INTRODUCTION (RTM) is a composite molding process which is widely used due to its ability to form net-shape structures outside the autoclave. In RTM, after one designs the mold cavity in the shape of the structure to be manufactured, the fiber preform that serves as the reinforcement is placed in the mold. The mold is closed and a viscous thermoset resin is injected into the cavity through an opening or openings referred to as gates to fill all empty spaces between the fibers and saturate the preform. The injection is discontinued shortly after the resin reaches a vent. Ideally, the vent hole should be placed at the location where the resin will arrive last. However, if the resin arrives at the vent before saturating all regions, the regions where fibers are not covered by resin are called dry spots. Dry spots are manufacturing defects and will result in rejection of the part. RTM has the potential for high yield, high performance and low cost manufacturing provided that it can be automated effectively. One of the key challenges is the uncertainties associated with the architecture and the placement of the preform [1,2]. A small gap between the mold wall and the reinforcement, or different fiber architecture along the preform edges will result in a phenomenon called race-tracking, in which the resin runs faster along the edges during the resin impregnation stage [3,4]. The degree of race-tracking will be a function of many parameters such as operator skill, variability in preform architecture, stacking and placement of preforms and process parameters. Hence it is difficult to reproduce from one experiment to the next. All these variations or uncertainties make it difficult to foresee the flow front pattern and therefore the final dry spot location, which is likely to change during each injection cycle making it difficult to ensure no dry regions during a production run. Different approaches have been developed to address this issue of race-tracking variation and dry spot formation. Most of them involve the use of process models and simulations along with sensing and control [5–12]. However, in order to use simulation tools accurately to predict and control the flow, one needs to input permeability values for the preform and account for the location and strength of race-tracking during mold filling to predict last regions to fill or to displace them. The present study introduces an approach to measure the bulk permeability and identify the location and strength of race-tracking for any mold configuration. The approach requires one to create a database using mold filling simulations that will contain the resin arrival times at every possible sensor location in the mold for all permutation of race-tracking strength along the edges of the mold cavity. The next step would be to place flow arrival sensors in the mold cavity at strategic locations and record the time of resin arrival during the injection phase. Finally by comparing the resin arrival times in the mold with the generated database, one could identify the closest scenario in progress during impregnation and also estimate the bulk permeability for that particular injection. This information could prove useful to execute on-line control, once all input permeabilities and race-tracking information is known. It could also be useful in identifying the important variables that cause race-tracking. After discussing previous efforts, we introduce our methodology and present a case study to validate our approach.
R
ESIN TRANSFER MOLDING
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
On-line Characterization of Bulk Permeability and Race-tracking
1527
ON-LINE PERMEABILITY MEASUREMENTS Here on-line refers to during the progression of the manufacturing process as opposed to off-line, which implies not during the process. Bulk permeability characterization has always been a major concern for designers. Simulation tools have helped them to carry out investigations on computers instead of the trial and error approach on the manufacturing floor. Furthermore modeling and predicting the filling behavior of RTM process is a necessary step for process automation. Simulation tools will reproduce the real world to the extent of the accuracy of parameters such as bulk permeability or race-tracking location and strength used in the simulations. Much research has been carried out with the aim of developing procedures and methodologies for bulk permeability measurements. Most of them are derived from the empirical Darcy’s law and provide the user with permeability values. Usually the experiment is performed separately (off-line) in a long wide mold creating one dimensional flow, see [13–18] for details. However bulk permeability depends on fiber volume fraction, and preform architecture; any changes in these during the placement and when it conforms to the net shape geometry of the mold cavity are not accounted for during impregnation or simulations. This could cause the resin to change its impregnation history and arrive at the vents before complete saturation of the preform causing dry regions. Measurement of bulk permeability in the mold geometry will allow one to create a database which could be used to improve the process, reduce the variations in preform architecture and provide guidance to preform manufacturers for tolerances. Research has also been conducted to develop methods to characterize race-tracking along different geometries [19–22] and to identify race-tracking with the help of sensors [5–12]. This research goes a step further and allows the user to identify the strength and location of race-tracking on-line along with values of preform bulk permeability.
Creation of Detection Database The approach to estimate race-tracking and bulk permeability can be described in two steps. The first step is off-line and generates simulations that map the potential range of race-tracking strengths. Note that the strength (or relative permeability) of a race-tracking region Ri is defined as i, the ratio between the permeability KRTi along the channel resulting in the race-tracking effect and the bulk permeability in the race-tracking direction. i ¼
KRTi Kbulk
The race-tracking strength domain is bounded by the minimum and maximum values, which can be assessed by examining the material used as well as the equipment available to cut the preforms at one’s disposal. Furthermore, research has been conducted to develop an approach to identify the race-tracking strength as a function of the size of the channel between the fabrics and the wall [3], which can help to assess these bounds. The physical dimensions of the gap that leads to race-tracking must be estimated to discretize the geometry for simulations so that one can examine scenarios with fixed race-tracking strengths [23,24]. In this study, we selected 5 discrete values listed in Table 1, along with possible race-tracking strength with the gap. The physical size of the gap is not as crucial in the simulation as the value of the strength in influencing the flow front pattern.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1528
M. DEVILLARD
ET AL.
Table 1. Race-tracking strength, q, which is a ratio of the increase in permeability along the edge as compared to the bulk value, is described for a typical edge and perform. Strength ()
Description
Gap [3] (mm)
3 7 14 32 70
Very low RT, almost unavoidable Low RT Medium RT Strong RT Very strong RT
0.02 0.1 0.5 1 1.6
R2 R4 R5
R3
Injection line
R1 Figure 1. Selected part geometry along with possible race-tracking regions.
Once the race-tracking domain is characterized, one must identify the regions where race-tracking is likely to occur. This can be achieved by examining the part geometry. The one analyzed in this work is shown in Figure 1. Race-tracking is likely to occur along most edges because of the uncertainties associated with the architecture and the placement of the preform in the mold. At this stage, one identifies a discrete number of disturbances resulting from the permutation of the race-tracking strengths (discrete values) over the race-tracking regions. One can execute simulations to generate flow front patterns for every permutation of the selected disturbances. Hence the total number of different simulations and flow front pattern created will be Np, where N is the number of discrete values of race-tracking selected and p is the number of regions in the mold where race-tracking is likely to occur. Each simulation showing a different combination of race-tracking strength will be referred to as scenario. In our case study, 3125 scenarios (N ¼ 5 and p ¼ 5) were created to map all disturbances likely to occur during injection. Simulations were executed using Liquid Injection Molding Simulation (LIMS), which was developed at the University of Delaware. LIMS is a finite element/control volume solver for the fluid flow through the porous media in a LCM process. It takes advantage of appropriate algorithms to decrease significantly the simulation time [25,26]. To estimate race-tracking, sensors need to be placed at strategic locations to identify each of the 3125 scenarios. A genetic algorithm referred to as SLIC [27] was used to provide seven sensor locations that could identify any selected scenario based on the arrival of the resin at these locations. The number of sensors was optimized such way that it is the minimal number of sensors required to distinguish all disturbance scenarios uniquely. Additional sensors do not necessarily improve the robustness of the system. The optimal locations are found by minimizing a cost function that maximizes the arrival time difference for the resin at these locations for different scenarios.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
On-line Characterization of Bulk Permeability and Race-tracking
1529
Figure 2. The flow front pattern is a function of , ratio of the permeability of race-tracking channel and bulk permeability. A change in relative permeability will affect the filling time; but for the same , the flow front pattern is identical.
In order to execute simulations, absolute bulk permeability and race-tracking channel permeability needs to be provided as input. However, through scaling one can show that the filling patterns are influenced only by the ratio of the two permeabilities. Therefore, the bulk permeability can be set arbitrarily even though it does not match the real permeability of the preform. Thus, as long as the relative permeability values are identical, simulations that use different bulk permeability values will generate the same flow front pattern; only the time scale will differ. Figure 2 shows an example to support the time scaling argument. Hence, one must also normalize the times of resin arrival at the selected sensor locations to distinguish the Np scenarios. This is accomplished by scaling the arrival times with the sum of all the arrival times to remove its dependence on the choice of absolute bulk permeability. All the normalized arrival times for each scenario are stored in a file for comparison with the on-line experiment in which the bulk permeability and the racetracking strength are unknowns. The methodology determines them by collecting the resin arrival times at the selected sensor locations. The simulations are used with predefined disturbances to predict flow patterns for each scenario and genetic algorithms are employed to find optimal sensor locations to uniquely identify each scenario. The uses of genetic algorithms and general search techniques have contributed to process optimization [28–30]. The table with normalized arrival times is created following the procedure illustrated in Figure 3 for on-line comparison with experiments in which all permeabilities are unknowns.
Race-tracking Detection in Experiments Once the detection database is generated, one can use it as a master reference chart to detect the scenario in an experiment using only the information from resin arrival times at the sensors. The mold cavity for which the database was generated was formed by placing an L-shaped insert between two parallel plates. Preforms were cut and placed around the insert in the mold geometry and point sensors that could detect resin arrival times were placed at the selected locations.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1530
M. DEVILLARD
ET AL.
R2 R4
R5 R1 N regions (R1, R2, ..), where race-tracking is likely to occur
Injection gate
-
Bulk permeability is arbitrarily selected
Discretize the race-tracking strength domain for each racetracking region. Using p discrete values
Simulate resin flow for Np scenarios.
Use genetic algorithms to find optimal sensor locations to uniquely identify each of the Np scenarios
R2 R5
R4 R3 R1
Normalize times for each scenario and create detection table
Figure 3. Procedure to create detection table for a mold geometry in which the regions and the strengths of race-tracking are specified.
Once the mold was closed, resin was injected at the selected gates as in the simulation. A lab-view application was created to collect the resin arrival times by the sensors during the experiment. These times were normalized and compared with the database to identify the closest of the Np scenarios in the database that matches the experiment in progress. The identified scenario, which shows the lowest least square error for the arrival times at all sensor locations, should feature the same filling pattern as in the experiment. However, the filling time may be significantly different. From the identified scenario, one not only knows the race-tracking location but also their relative permeability values for the on-line experiment since those relative permeabilities will be similar to the input values used to obtain the identified simulation. The limitation of this methodology is that we assume that the race-tracking strength is uniform along a given edge, which is not always true in practice. However, previous work showed that a nonuniform race-tracking can be accurately modeled by a uniform race-tracking associated with an average strength [24]. Also, in the experiments there may be variations in the bulk permeability which will lead to results that could not exactly match the predefined scenarios.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1531
On-line Characterization of Bulk Permeability and Race-tracking
Bulk Permeability Characterization The scaling analysis of the Darcy’s law (1) [31–33] performed below allows us to derive a relationship between the simulated filling time, the experimental filling time, the arbitrary bulk permeability used to scale the race-tracking permeability and characterize the permeability of the material in the mold. u ¼
K rP
ð1aÞ
or in the 2D form assuming that Kxy ¼ 0,
1 Kxx @x=@t ¼ @y=@t 0
0 Kyy
@P=@x @P=@y
ð1bÞ
Here ux ¼ ð@x=@tÞ and u y ¼ ð@y=@tÞ, where ux and uy are components of Darcy velocity u , in x and y directions respectively. By normalizing properly each variable, one obtains, L @x^ Kxx Pc @P^ ¼ L @x^ tc @t^
ð2aÞ
L @y^ Kyy Pc @P^ ¼ tc @t^ L @y^
ð2bÞ
@x^ Kxx Pc tc @P^ ¼ L2 @x^ @t^
ð3aÞ
Kyy Pc tc @P^ @y^ ¼ L2 @y^ @t^
ð3bÞ
u^ ¼ ½K rP^
ð3cÞ
and
which leads to
and
or
Therefore by stating that (3c) holds for the experiment and the simulation, we obtain for the x variable, Kxx Pc tc Kxx Pc tc ¼ ð4aÞ L2 exp L2 sim
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1532
M. DEVILLARD
ET AL.
which can be recast as follows,
½Kxx exp
2 ðtc Þsim ðPc Þsim L exp ¼ ½Kxx sim ðtc Þexp ðPc Þexp ðL2 Þsim
ð4bÞ
Note that (Pc) and L, which are respectively the inlet pressure and the distance used for normalizing x and y, have the same values for the experiment and the simulation. Therefore
½Kxx exp ¼
ðtc Þsim ðÞexp ½Kxx sim ðtc Þexp ðÞsim
ð5Þ
Similar scaling exists between [Kyy]exp and [Kyy]sim. Here [Kxx]exp is the bulk permeability of the material used during the experiment in the direction, [Kxx]sim is the arbitrarily chosen bulk permeability in the simulation. sim and exp are the viscosities of the resin used in the simulation and the experiment respectively. tcsim is the sum of the arrival times at each virtual sensor location of the selected scenario and tcexp is the sum of the arrival times at each sensor location in the mold during the experiment. This scaling analysis and especially the relation (5) provide a relationship between the experimental bulk permeability and the arbitrary bulk permeability set for simulation purposes. During the experiment, one registers the times when the resin arrives at the sensors and this information is compared with the detection table to identify the scenario in progress, the strengths of race-tracking channels and also the bulk permeabilities of the preform using Equation (5). Figure 4 illustrates procedure for the on-line experiment. mold Labview code sensors Experiment
Detection database.
Collects and records times at which sensors are triggered, computes the normalized times and find a match within scenarios available in the database. Race-tracking identification Using relation (5), characterize bulk permeability.
Figure 4. Schematic of the on-line procedure.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1533
On-line Characterization of Bulk Permeability and Race-tracking
The knowledge of the bulk permeability and the race-tracking strength of all race-tracking regions provide the complete permeability map of the experiment in progress that can be used for flow monitoring or control.
CASE STUDY The methodology is validated here with the help of a case study.
Experimental Setup The part geometry is shown in Figure 5 along with injection gates, vents and detection sensors. In order to achieve a fully automated system, a supervisory computer controls all features. Furthermore with the aim of validating the ability of the system to identify the appropriate simulation scenario, we also recorded the experiment with a CCD camera through the 2-in. thick acrylic top plate in order to compare simulated and experimental flow front patterns. To obtain feedback about resin arrival during injection, sensors have been placed into the mold. Their locations were selected through the off-line optimization. Point sensors are integrated into the tooling and connected to the data acquisition system with the appropriate National instrument hardware. As shown in Figure 6, point sensors consist of (a) two twisted electrical wires, which are held into the bottom plate of the mold by glass fibers and (b) cured resin. Once the resin is cured, (c) the extra cured resin is removed. The presence of injected liquid is determined by an increase in current flow when the liquid wets the two poles of the sensors reducing the resistance between the two leads. The current is detected by a half-bridge excited by a 10 VDC power supply. The resistor through which the current is detected must be adjusted to the type of fluid used. In this study a blend of water and corn syrup mixed at the suitable ratio to obtain the desired viscosity was used as highly conductive working fluid. A resistor of 325 was used for corn syrup-based working fluid. One will need to increase Video camera R2
Pressurized bucket containing the liquid to be injected
R3R4
R5
Injection line R1
Point sensors
Vent
(a)
L-shaped mold
(b)
Figure 5. (a) Part geometry and sensor locations; (b) Experimental setup.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1534
M. DEVILLARD (a)
(b)
ET AL.
(c)
Twisted wires
reinforcement
Figure 6. Procedure to fabricate a point sensor to detect resin arrival at that location.
the resistor up to 75 M to detect vinyl-ester thermoset resin arrival at the point sensor since common vinyl-ester thermoset resins are less conductive. Dielectric point sensors are widely used for their reliability. Nevertheless point sensors may affect surface quality and be completely ineffective as one use gel coat, which covers up sensor tips, or carbon fibers, which are conducting. Therefore, industrial parties might be reluctant to use as many point sensors as required or constrain the location where point sensors can be placed. To address those issues promising sensors based on magnetic field, referred to as TDR are currently being developed at the University of Delaware/CCM and could be implemented to the methodology described in this paper. TDR sensors are placed below the surface of the molding tool, which do not affect the surface quality [34].
Experiments A set of experiments were conducted to validate this methodology. Two different glass fiber materials were used, (i) Woven roving (Vetrotex I3BN324500) and (ii) Random mat roving (Vetrotex). The resin viscosities used were 120 and 220 cp. Five operators took part in this project in order to study the variability of the racetracking strength. Four of them (Operators A,B,C,D) cut the fabrics by hand and placed them into the mold before the experiment was initiated. The last operator E cut the fabrics with a laser cutter before placing them into a mold. In the next section, we highlight the ability of the system to determine along which edges race-tracking occurs as well as its strength. Furthermore we study how these parameters vary as a function of the operator. Then we determine the bulk permeability of the preform using the information gathered by the point sensors. More than 50 experiments were conducted to obtain statistically significant results.
Race-tracking Identification All results were obtained during the progression of the mold filling stage. After all the sensors were triggered the sequence of the triggering was compared with the detection database and the possible scenario in progress was identified. To validate if the correct scenario was being identified with correct values of race-tracking strengths, the flow front
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
On-line Characterization of Bulk Permeability and Race-tracking
1535
progression during the experiment was compared with the identified simulation. If experimental and simulated flow front patterns matched, one could safely conclude that the relative permeability values for the experiment are similar to the ones used for the simulation whose arrival times at the sensor locations matched the experiment. The overall system was able to select the most appropriate scenario on-line (scenario showing a flow front pattern identical to the experiment). This was validated by comparing the movie of the experiment recorded through the acrylic top lid and the simulation that was selected. If the flow front patterns matched, the methodology of racetracking identification was successful. Fifty-three experiments were conducted and the flow fronts matched within a few millimeters at all times in fifty experiments, thus validating identification of the correct scenario. For the remaining experiments, the selected scenario and experiment were visually different. This is understandable considering the limitation of the assumption of uniform race-tracking and bulk permeability, which may have varied in some of the experiments. Figure 7 shows an example of one such comparison. As the methodology was able to identify in each experiment the strength of racetracking along the susceptible edges, we decided to use this information to study the influence of operator on race-tracking strength. The operator task consisted of cutting fabrics by hands for Operators A, B, C and D and with a laser cutter by Operator E and placing fabrics into the mold. The results shown in Figure 8 are the data on race-tracking strength for 50 experiments conducted with the woven preform with five different operators (10 experiments by each operator). This histogram, shown in Figure 8, was obtained by finding the average race-tracking value in all 10 experiments conducted by each operator and normalizing it by the highest race-tracking strength at each edge. The possible strength values are listed in Table 1. Some extra pieces of information are also displayed on this chart such as the average strength for a region, which highlights the strength variation from one race-tracking region to another. The race-tracking strength varied depending on the skill of the operator to cut and place the preform. Furthermore some regions appear to be more likely to cause race-tracking than others, Regions R1 and R2 show the highest tendency , this is consistent with the fact that the preform along R1 and R2 is in contact with a very stiff aluminum frame , while along R2 and R3 the preform is in contact with an insert made of rubber. The standard deviation of race-tracking strength for each operator also provides valuable data that can be used to describe the variation of race-tracking strength depending on the skill of the operator (Figure 9). Higher the standard deviation, the less consistent was the operator in placement of the preform along that edge or/and less was the tolerance in the fabric architecture. This study confirms that not only race-tracking exists but its strength is unpredictable and nonrepeatable from one experiment to the next. Operator care or the use of laser cutter can narrow the range but it will not make flow patterns identical or repeatable. A process designer must account for these issues, when selecting injection and vent locations. Research is being conducted to address this issue [5–12].
Bulk Permeability Characterization In addition to identifying race-tracking, one can also use the sensor times to estimate the bulk permeability of the preform during the impregnation experiment. Permeability
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1536
M. DEVILLARD
ET AL.
time
Scenario 743 was selected Figure 7. Comparison of flow fronts between the experiment (Kxx ¼ 3 10 10 m2) and the identified simulation (Kxx ¼ 4 10 12 m2).
measurements and characterization are usually carried out in a separate one, two or three dimensional mold and it is a challenge to measure consistent permeability values [13–18]. Twenty to thirty percent variations in measurements are not uncommon. Our methodology allows one to find the permeability of the preform during the mold filling stage in the mold cavity of the part being manufactured. This information can allow one to create a statistical database on the tolerance of the preform type and its architecture and also be useful for controlling the flow.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1537
On-line Characterization of Bulk Permeability and Race-tracking 80
70
race-tracking strength
60
Average R-1 23.2
Average R-3 22
50
40
Average R-2 15.3
30
Average R-4 7.2
Average R-5 5.5
R 4A R 4B R 4R R C 4- 4 E - -D La se r
R 5A R 5B R 5R R C 5- 5 E - -D La se r
20
10
R 3A R 3B R 3R R C 3- 3 E - -D La se r
R 2A R 2B R 2R R C 2- 2 E - -D La se r
R 1A R 1B R 1R R C 1- 1 E - -D La se r
0
Figure 8. Race-tracking strength distribution by regions and operators. 120
100
standrad deviation
80
60
40
20
5A R 5B R 5C R R 5- 5E- D La se r
R
4A R 4B R 4C R R 4- 4E- D La se r
R
3A 3B R 3C R R 3- 3E- D La se r R
R
2A 2B R 2C R R 2- 2E- D La se r
R
R
R
R
1A 1B R 1C R R 1- 1E- D La se r
0
Figure 9. Standard deviation in race-tracking strength by operators.
In order to validate the method, we used three different detection databases with three different bulk permeability values of Kxx as 3 10 10, 4 10 12 and 7 10 8 m2. Using Equation (5), we determined the experimental Kxx value using the times of resin arrival at the sensor locations. As expected, the use of any of the above permeability values in the
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1538
M. DEVILLARD
ET AL.
K4 9
K4 7
K4 5
K4 3
K4 1
K3 9
K3 7
K3 5
K3 3
K3 1
K2 9
K2 7
K2 5
K2 3
K2 1
K1 9
K1 7
K1 5
K1 3
K9
K1 1
K7
K5
K3
K1
Kxx m2
simulation resulted in the same value for the Kxx for that experiment. However, the Kxx value varied from one experiment to the next. The values determined for all fifty experiments are shown in a histogram shown in Figure 10 for woven material and Figure 11 for random mat material. The average value obtained over 50 experiments is 2.7 10 10 m2, with a standard deviation of 1.02 10 10 m2. The permeability of this preform was measured inside a 1D mold, the value was found to be as 3 10 10 m2 [35]. The variation of bulk permeability is to be expected due to the variation in the preform architecture, the lay-up and the inability to accommodate the same fiber volume fraction in every experiment. The anisotropy of this material was found to be 1.23, which is again consistent with one dimensional experiments used to measure permeability in x and y direction separately. We also conducted experiments with an isotropic random mat. All preforms were cut with a laser cutter and the viscosity of the impregnating fluid was 120 cp for five experiments and 220 cp for the last five. Figure 10 shows the outcomes computed from the detection database, in which the bulk permeability used was 4 10 12 m2. Again the outcomes were identical irrespectively of the value used for the simulation due to the use of the Equation (5). This was confirmed by using 3 10 10 and 7 10 8 m2 as two other bulk permeability values in the simulation, which resulted in the same values for the experiments shown in Figure 10. The average value over the ten experiments with random preform was found to be 1.47 10 9 m2. The standard deviation was 5.32 10 10 m2. One can note that this perform permeability does have much lower standard deviation than four by five woven preform due to a more uniform structure. As expected, the average permeability value for random mat is higher than for woven material since fiber tows in the woven material offer
Figure 10. Histogram of measured on-line permeability value in the x direction for fifty experiments using 4 5 woven preform. The computed average value for this woven material was 2.7 10 10 m2 and the standard deviation was 1.02 10 10 m2.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1539
On-line Characterization of Bulk Permeability and Race-tracking 2.5E-09
2.0E-09
1.5E-09 Kxx m2
1.0E-09
5.0E-10
0.0E+00 K1
K2
K3
K4
K5
K6
K7
K8
K9
K10
experiments
Resin viscosity: 120cp
Resin viscosity: 220cp
Figure 11. Histogram of measured on-line permeability value in the x direction for ten experiments for random preform.
more resistance to the flow. Also Equation (5) takes into account viscosity effect, which does not affect our methodology as it can be seen in Figure 11.
CONCLUSION The characterization of race-tracking and the measurement of bulk permeability are two key issues in the understanding and control of resin impregnation step during resin transfer molding process. This work focused on the identification of race-tracking and the measurement of its strength along with the measurement of the bulk permeability during the impregnation step. The procedure outlined in this paper could be ideally used to handle any complex part for RTM or VARTM. The approach uses resin impregnation simulation tools with predefined disturbances and optimized sensor locations to identify the location and strength of race-tracking. A database was created by permutations of selected mold edges and race-tracking strengths. By comparing the sensor arrival times in the experiment with the created database, the relative strength and location of race-tracking was identified. By proper scaling arguments one could also measure the bulk permeability of the preform during the experiment. A case study was presented to illustrate the reliability of the approach and the variation in the bulk permeability values. The variability in preform permeability and race-tracking can result in flaws for composite materials manufactured by this process such as dry regions, hence the need to address them during the mold and process design stage.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
1540
M. DEVILLARD
ET AL.
ACKNOWLEDGMENT The authors gratefully acknowledge the support provided by the Office of Naval Research (ONR) under Grant #N00014-97-C-0415 for the ‘Advanced Materials Intelligent Processing Center’ at the University of Delaware. Additional appreciation goes to the operators involved in this project, Estrada Gonzalo, Laut Antoine, Millas Carlos.
REFERENCES 1. Advani, S.G., Bruschke, M.V. and Parnas, R. (1994). Chapter 12: Resin Transfer Molding, In: S.G. Advani (ed.), Flow and Rheology in Polymeric Composites Manufacturing, pp. 465–516, Elsevier Publishers, Amsterdam. 2. Busch, J. (1998). Composites Molding Processes, Advanced Materials & Processes, 153(3): 25–26. 3. Bickerton, S. and Advani, S.G. (1999). Characterization and Modeling of Race-tracking in Liquid Composite Molding Processes, Composites Science and Technology, 59(15): 2215–2229. 4. Han, K., Wu, C.H. and Lee, J. (1993). Characterization and Simulation for RTM Race-tracking and Dry Spot Formation, In: Proceedings of Ninth Annual ASM-ESD. 5. Lawrence, J., Hsiao, K.-T., Simacek, P. and Advani S.G. (2002). An Approach to Couple Mold Design and On-line Control to Manufacture Complex Composites Parts by Resin Transfer Molding, Composites, Part A: Applied Science and Manufacturing, 33: 981–990. 6. Luo, J., Zhang, Z.L. and Wang, B. (2001). Optimum Tooling Design for Resin Transfer Molding with Virtual Manufacturing and Artificial Intelligence, Composites Part A: Applied Science and Manufacturing, 32(6): 877–888. 7. Nielsen, D. and Pitchumani, R. (2001). Intelligent Model-based Control of Preform Permeation in Liquid Composite Molding Processes, with Online Optimization, Composites Part A: Applied Science and Manufacturing, 32(12): 1789–1803. 8. Demirci, H., Hamdi, H. and John P. Coulter (1994). Neural Network Based Control of Molding Processes, Journal of Materials Proceedings and Manufacturing Science, 6: 335–354. 9. Demirci, H., Hamdi, H. and John P. Coulter (1995). Control of Flow Progression During Molding Processes, Journal of Materials Processing and Manufacturing Science, 7: 409–425. 10. Mogavero, J., Sun, J.Q. and Advani, S.G. (1997). A Nonlinear Control Method for Resin Transfer Molding, Polymer Composites, 18(41): 2–7. 11. Lunstorm, S. and Gebart, R. (1994). Influence from Process Parameters on Void Formation in RTM, Polymer Composites, 20(4): 25–33. 12. Beker, B., Barooah, P., Yoon, M.K. and Sun, J.Q. (1994). Sensor Based Modeling and Control of Fluid Flow in RTM, Journal of material Processing and Manufacturing Science, 7(2): 161–170. 13. Wang, T.J., Wu, C.H. and Lee, L.J. (1994). In-plane Permeability Measurement and Analysis in Liquid Composite Molding, Polymer Composites, 15(4): 134–142. 14. Weitzenbock, J.R., Shenoi, R.A. and Wilson, P.A. (1999). Radial Flow Permeability Measurement, Composites: Part A, 33: 781–796. 15. Nedanov, P. and Advani, S.G. (2002). A Method to Determine 3D Permeability of Fibrous Reinforcements, Journal of Composite Materials, 36: 241–254. 16. Luo, Y., Verpooest, I., Hoes, K., Vanheule, M. and Cardon, A. (2001). Permeability Measurement of Textile Reinforcements with Several Test Fluids, Composites Part A: 23: 1497–1504. 17. Chan, A.W. and Morgan, R.J. (1997). Modeling Race Tracking Effects in Liquid Composite Molding, CAE and Intelligent Processing of Polymeric Materials, AMSE, Materials Division, 79: 361–367. 18. Robitaille, F., Long, A.C., Rudd, C.D. and Souter, B.J. (1998). Measurement of In-plane Permeability for Sheared Preforms, In: Proceedings of the International Conference on Computer Methods in Composite Materials, pp. 495–504.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
On-line Characterization of Bulk Permeability and Race-tracking
1541
19. Laval, N., Sozer, E. and Advani, S.G. (2000). Characterization of Permeability Around a 90
Corner, Advanced Composite Letters, 17–24. 20. Bickerton, S., Sozer, E., Simacek, P. and Advani, S.G. (2000). Fabric Structure and Mold Curvature Effects on Preform Permeability and Mold Filling in the RTM Process, Composites Part A: Applied Science and Manufacturing, 93(10): 439–458. 21. Ramakrishnan, B. and Pitchumani, R. (2000). Fractal Permeation Characteristics of Preforms Used in Liquid Composite Molding, Polymer Composites, 21: 281–296. 22. Woerdeman, D.L., Phelan, F.R. and Parnas, R.S. (1999). Interpretation of 3D Permeability Measurements for RTM Modeling, Polymer Composites, 470–480. 23. Gokce, A. and Advani, S.G. (2001). Combinatorial Search to Optimize Vent Locations in the Presence of Disturbances in Liquid Composite Molding Processes, Submitted to Materials and Manufacturing Processes. 24. Gokce, A. and Advani, S.G. (2003). Race Tracking Forecast in Liquid Composite Molding Processes Case Study: Vent Optimization to Minimize Dry Spot Formation, SAMPE Symposium and Exhibition, 19(4). 25. Simacek, P. and Advani, S.G. (2002). LIMS 5.0 and LIMS UI User Manual, Tech. Report, CCM, University of Delaware. 26. Maier, R.S., Rohaly, T.F., Advani, S.G. and Fickie, K.D. (1996). A Fast Numerical Method for Isothermal Resin Transfer Mold Filling, International Journal of Numerical Methods in Engineering, 39: 1405–1422. 27. Hsiao, K.-T., Devillard, M. and Advani, S.G. (2002). Streamlined Intelligent RTM Processing: From Design to Automation, In: Proceedings of 47th International SAMPE Symposium and Exhibition, 47: 454–465. 28. Gokce, A., Hsiao, K.-T. and Advani, S. G. (2001). A Method to Find Auxiliary Injection Gate Locations for Successful Mold Filling in Resin Transfer Molding Process, In: Proceedings of 46th International SAMPE Symposium and Exhibition, 46: 310–325. 29. Lin, M. and Hahn, H.T. (2000). Resin Transfer Molding Process Optimization, Composites, Part A: Applied Science and Manufacturing, 31: 361–371. 30. Mathur, R., Advani, S.G. and Fink, B.K. (1999). Use of Genetic Algorithms to Optimize Gate and Vent Locations for the RTM Process, Polymer Composites, 20(2). 31. Advani, S.G. and Bruschke, M.V. (1994). A Numerical Approach to Model Non-isothermal, Viscous Flow with Free Surfaces Through Fibrous Media, Methods Fluids, 19: 575–603. 32. Fracchia, C.A., Castro, J. and Tucker III C.L. (1995). A Finite Element/Control Volume Simulation of RTM Filling, In: Proceedings of the American Society for Composites Fourth Annual Technical Conference. 33. Advani, S.G. and Bruschke, M.V. (1990). A Finite Element/Control Volume Approach to Mold Filling in Anisotropic Porous Media, Polymer Composites, 11(6): 398–405. 34. Dominauskas, A., Heider, D. and Gillespie, J.W. (2003). Electric Time Domain Reflectometry Sensor for On-line Flow Sensing in Liquid Composite Molding Processing, Composites Part A: Applied Science and Manufacturing, 34(1): 67–74. 35. Estrada, G., Celine Vieux-Pernon and Advani, S.G. (2002). Experimental Characterization of the Influence of Tackifier Material on Preform Permeability, Composites Materials, 36(19): 2297–2310.
Downloaded from http://jcm.sagepub.com at PENNSYLVANIA STATE UNIV on April 15, 2008 © 2003 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.